Hi Ibrahim,
I assume you are talking about MPLS-based Service Provider (SP) networks.
Well, a provider's backbone does not consist just of edge devices, right? It must also have some internal routers that provide connections between the edge devices. Of course you have the PE routers being the routers that connect individual customers to the provider's network. But the provider's network would need to be quite small and fully meshed if it was to consist just of PE routers. Clearly, that is out of question.
So if there are any intermediate routers between the PE routers that do not connect to any customer devices directly, these routers are by definition P routers. We distinguish between PE and P routers so diligently because they really serve different purposes and they also have different knowledge about the surrounding network. PE routers are required to know both SP's network and the network of each customer they connect to. P routers are supposed to know only the SP's network - this is a huge advantage of MPLS-based networks. The P routers do not need to be aware of any customers. As a result, the core of SP's network is not influenced by the amount of customers. This is a huge scalability asset and advantage. P routers deliver packets based on MPLS labels rather than IP headers, but the MPLS label have to be initially set by the PE routers.
I know this is a brief answer but the P routers are simply the routers inside the provider's backbone whose main purpose is to interconnect PE routers. It's as simple as that.
Please feel welcome to ask further!
Best regards,
Peter
